For most of history, knowledge was the thing that separated the top of an organization from the bottom. The senior person wasn't senior because they worked harder. They were senior because they'd seen more, made more mistakes, and accumulated a body of judgment the junior person hadn't earned yet. The whole structure of a company, a guild, a hospital, an army, ran on that gap. You climbed the chain by slowly closing the distance between what you knew and what the person above you knew. Execution was the reward for having enough knowledge to be trusted with it.

That gap is what's actually closing right now, not jobs, the gap itself. Anthropic released Claude Fable 5.1 on September 1, 2026, built specifically to run long, unattended coding and research sessions, following a plan across hours without a human walking it through every step. A few weeks before that, OpenAI previewed a model internally called Astra to US senators, a multi agent system where the agents split up a long task and execute it together instead of a person managing each piece by hand. Different labs, different names, same direction. Both are less about answering questions and more about carrying out a plan on their own once you've told them what you want.

That's the part worth sitting with. It's not that these models know more than a junior employee. In a lot of narrow ways they know less, they don't have your company's context, your market's quirks, your users' actual behavior. What they've closed is the distance between having an idea and having it done. You used to need years of technical depth before anyone would let you touch execution. Now the technical layer is thinning out fast, and what's left standing is whether you can think clearly enough to write down what you actually want built.

Which is why a good PRD is starting to matter more than a strong resume of years spent in the weeds. Not because thinking is easy, it's the opposite, it's the only thing left that's hard. The part that used to be hard, translating a plan into working execution, is the part getting automated first. The part that stays hard is knowing what's worth building in the first place, reading a market correctly, understanding what a user actually needs versus what they say they need, and being precise enough in your own head that a model can act on your instructions without you babysitting every step.

I don't think this makes the experience worthless. I think it makes a specific kind of experience worthless, the kind that was really just "I've done this exact repetitive task enough times to be fast at it." That kind was always going to be the first thing automated, in every field, not just software. What survives is the kind of experience that shows up as judgment, knowing which problems are worth solving, spotting a bad plan before it's executed, holding a clear model of a market in your head. That's harder to fake and harder to shortcut, and it's exactly the skill that's about to get more valuable while the technical execution layer underneath it gets cheaper.

What I'm actually curious about is what happens to the next generation of jobs once this plays out fully. If the barrier to acting drops close to zero, then the people who used to compete on having more technical knowledge than everyone else lose that edge fast. What's left is a much smaller pool of people who are good at deciding what to act on, and a much larger pool of people who can execute anything as long as someone tells them exactly what to do. I don't know yet which side of that line most careers will end up on. But it's clear which side is getting more crowded, and it's not the thinking side.